Looks like the final tally will be in the neighborhood of 18 to 20 points. Marc Ambinder says:
The rift in the party widens: Obama voters by and large would NOT be satisfied (55%) with Clinton as their nominee, while 7 and ten Clinton voters would NOT be satisfied (72%) with Obama as theirs.
It's not accurate to say that the rift is widening. The results in neighboring states produced almost exactly the same figures. In Louisiana, 54 percent of Obama voters said they wouldn't be satisfied with Clinton, and 70 percent of Clinton voters said they wouldn't be satisfied with Obama. In Alabama, the figures were 56 percent and 76 percent, respectively. Presumably this can be attrituted partly to race. (The figures are lower in states outside the Deep South, but not dramatically lower--pretty much everywhere, majorities of both Obama and Clinton voters have said they wouldn't be satisfied with the other candidate as the nominee. It's possible to read too much into these results: People like their candidate and want him/her to win.)
On another note, John King on CNN just pointed out that Clinton did better in more Republican parts of Mississippi, and implied that this might mean that Clinton would do better among Republicans in November. This is patently the wrong conclusion to draw, and is a good demonstration of the perils of attempting to predict general-election outcomes based on primary results. Obviously, whites voted heavily for Clinton, and white areas of the state support Republicans in presidential elections--thanks to the votes of people who don't vote in Democratic primaries. Clinton's strong showing in these parts of the state doesn't demonstrate anything about the preferences of a single Mississippi Republican. This is a simple enough point that CNN should be able to convey it to viewers.